Re: Very strange performance decrease when reusing a PreparedStatement

On Wed, 2009-05-06 at 18:39 -0400, Tom Lane wrote:
> Simon Riggs <simon@2ndQuadrant.com> writes:
> > We talked a few years back about an "Option node" that would select
> > sub-plans at runtime based upon the actual values of the bind variable.
> > Nobody's got around to doing anything with that yet.
>
> There is an AlternativeSubPlan mechanism in 8.4, though we're not doing
> very much with it yet.

OK, good. We just need to use it when the distribution of values is
highly non-uniform such that values with max and min selectivity can
generate different plans *and* we have a parameter, the arrival
distribution of which we have no way of knowing at plan time. That case
accounts for about 50% of the cases for which hints/special handling is
currently required.

--
 Simon Riggs           www.2ndQuadrant.com
 PostgreSQL Training, Services and Support